The Indo-European Resultative Particle *es

George E. Dunkel
Indogermanisches Seminar, Universität Zürich

Since I first met Jens in Erlangen in 1974 his flood of fearless, rigorously argued and re-
freshingly original solutions for tenacious linguistic problems has never ceased to delight
me. In this spirit I offer him this unabashedly speculative study of five notorious quandaries
in search of a solution. The first two of these occur before roots.

1. A preterital prefix *e- is found in Indo-Iranian, Phrygian, Greek and Armenian.

Anatolian may have either split off before the augment came into use or have lost it
along with most other dialects.1 If the syllabic augment was a feature of Late IE,2 it
may nonetheless have originated far earlier. But whether it was a parallel innovation of
the above dialects or is ascribed to Late IE, in either case this old and widespread
morpheme needs an explanation.
The IE literature unanimously asserts that *e- was originally a temporal adverb
which meant “in the past, then, olim”. This ubiquitous idea can however only be enter-
tained by those for whom the IE system of particles is of no interest whatsoever, since
it is not supported by any evidence at all.
Connecting the augment with Greek bebaiotic ἦ “truly, indeed” as proposed by M.
Bréal and K. Strunk (1994: 277) leaves its primary preterital function unaccounted
for.3
The same objection holds for the equation with the pronominal stem *e- (Strunk
1994: 277-9), as this is generally held to be not, with Strunk, far-deictic but rather ana-
phoric in function, i.e. a marker of “old information”. If the original function of the
augment had been anaphoric rather than far-deictic, it would have referred back not to
the past as such but only to an “aforementioned time”, this being indicated by a (usu-

1
The single -s- of Hittite ēšun need not (with Eichner, FT Regensburg: 78) reflect lenition, see Mel-
chert 1994: 152. I thank C. Melchert and S. Scarlata for helpful discussion; all errors are my
own.
2
Strunk 1994: 274 lists attempts to find traces in other IE dialects. A further possibility is the
reformation of augments into reduplications, e.g. Italic *efaked > fefaked, *etenet > tetinit, etc.
Against *eliquet > *leliquit > reliquit see my contribution to the forthcoming Festschrift Gren-
Eklund.
3
Strunk’s ascription of preteritality to Greek ἦ is in itself implausible and takes no account of the
traces of bebaiotic *ē in Latin and Vedic, e.g. deva- “devout” < delocutive “who says ‘ḗ dei̯u̯e’”
(cf. Lat. ēdī), see Syntaxe des langues indo-iraniennes anciennes, ed. E. Pirart 1997: 22-3.
90 George E. Dunkel

ally text-initial) frame-setting adverb4 or verb.5 Similarly, other adverbs and pronouns
with a seemingly optional initial e- (Greek ἐχθές, ἐκεῖνος, Oscan etanto and the like)
would also originally have referred back to their respective first occurences in the text.
However, since scene-setting expressions could just as well have been futuristic,6 the
augment’s exclusively preterital function would still remain unaccounted for.
2. *s-movable. A second pre-radical anomaly is the morpheme-initial alternation
by which an initial *s- alternates non-distinctively with zero,7 universally presumed to
occur only among verbal roots (but see the excursus after §3).
2a. The older approach was to assume a prehistoric gain of *s-. Usually movable *s-
is taken as a faded prefix; already A. Pott derived it from *som/s- “together” (Etym.
Forsch.2 I (1857): 290, 293ff.), an idea which is today indefensible.
Th. Siebs’ demonstration that IE initial voiced and voiced aspirated stops were
progressively (!) devoiced by the addition of a prefix *s- (Siebs 1904)8 presumed a
stage at which movable *s- was being actively added on. Siebs’ Law is evidently gram-
matically, not phonetically, conditioned; otherwise it would be impossible to recon-
struct the allophone [*z], found in 22 entries of Pokorny, Idg. etymol. Wörterbuch.9
In 1891 J. Schrijnen defined the original function of the faded *s- as intensive and
explicitly rejected the sentence-phonetic alternative which had arisen in the meantime
(§2b).10 Later he proposed that the intensive had arisen out of an earlier causative
function.11 Schrijnen eventually developed a whole theory of “preformants”: syn-
chronically the mirror-image of root-extensions and just as seemingly functionless,
diachronically the remnants of various faded prepositions or preverbs.12

4
*ĝʰz-dʰi̯-es “yesterday” (§9), *per-ut “last year”, *pr-u “in front of, preceding (the present), i.e. in
the past”. On personal vs. impersonal time-lines, where ego is or is not congruent with nunc, see
KZ 96 (1983): 67f., 78ff.
5
Whether the verbum substantivum *h₁es-se-t (Alkman 74 ἦσκέ τις Καφεὺς ἀνάσσων) or a form of
the root *bʰuh₂-, cf. Luvian puwa “formerly, karū” (V. Ivanov, Proceedings, 12th UCLA IE Confer-
ence (2001): 80ff.); cf. Russian žil-byl.
6
E.g. *h₁eseti “it shall happen” or a form of *apo-di̯eu̯- “the sky behind (the present one), i.e. the
future”.
7
See Mayrhofer 1986: 119-120, Szemerenyi 1996: 93-4, and Southern 1999.
8
Thus allowing comparisons such as Lat. fodio : Engl. spade; Lat. fūmus : Engl. steam; Latin fīvo/fīgo
: stīgo; Greek θιγεῖν : στίζω and so on.
9
Thus IE *mizdʰo- and Avestan zdī are irrelevant as counter-examples to Siebs’ law not because
the clusters are non-initial (Mayrhofer 1986: 92 fn. 13), but because they do not contain the pre-
fix. On *zges- see fn. 38.
10
Études sur le phénomène de l’s mobile (1891): 22, cf. KZ 38 (1905): 139. Siebs also rejected the sen-
tence-phonetic approach (1904:288-92) but held that the determination of the prefix’s original
meaning or function was impossible since it had already faded by IE times (1904: 292-3).
11
Scritti … Trombetti (1938) = Collectanea: 144-6.
12
KZ 42 (1908): 97ff. = Collectanea: 111ff.
 The Indo-European Resultative Particle *es 91

H. Hoenigswald applied the laryngeal theory to the question of movable s- before

vowels, certainly an embarassing gap for any theory of adventious s-.13 Leaving aside
uncertain material, comparisons like an-us – sen-ex, oc-ulus – seh-en, ἀρκέω – ἕρκος are
convincing enough to establish that initial laryngeals were lost between the prefix *s-
and vowel as Hoenigswald’s corollary to Siebs’ law.14 It follows that the prefixation
took place before the coloration of *e by laryngeals.
A different approach to adventitious *s- would be to take it not as an originally
meaningful particle but rather as a metanalysis of the common adverbial ending *-s
(fn. 24) by word-boundary shift, e.g. *ambʰis-pe-i̯o- > ambʰi-spe-i̯o-.15
2b. The more recent approach is antithetic: to assume a loss of etymological *s-.
The simplest conceivable mechanism, local dissimilation, is only rarely applicable, as in
*(z)ges- “quench” or *(s)tro-zd-o- “thrush”. We therefore turn to sentence-phonetic
factors.
In 1881 H. Osthoff postulated a loss of *s between stops in sequences like *ud
spe-, *tod spe- (while e.g. *peri spe- remained intact).16 However this was a parallel
innovation of Indic, Greek and other dialects rather than a feature of the proto-
language, as was already clear from the dialectal remnants of *-s- between stops (e.g.
Latin textus, posco etc.) even before the preservation of *-tst- in Hittite became known.
In 1885 J. v. Fierlinger pointed out that movable s- occurs before resonants as well
as before stops (KZ 27: 196 fn.). He therefore proposed that degemination after word-
final *-s (especially of nominatives) was the crucial mechanism, e.g. *nebʰos stenh₂i̯eti,
*i̯os stegeti both led to *-os te-. This has remained the favorite single explanation for s-
mobile. H. Hirt added particles in adverbial *-s (e.g. *eĝʰ-s, *ab-s or *ad-s) to the set of
degeminators (1927: 319f., 329-33) while F. Edgerton showed that such degemination
was still very much alive in Vedic (Language 34 (1958): 445-453).
2c. As there is undeniably much loss in the individual dialects17 degemination is
undeniably a plausible contributing factor, but Siebs and Hoenigswald have shown
that it cannot be the whole story, i.e. that addition must be involved as well. Still un-
clear remains whether the adventitious *s- was itself a particle, the remnant of a parti-
cle, or metanalysed from adverbial *-s. M. Southern 1999 applies the panoply of mod-
ern linguistics and concludes with a Hegelian synthesis through polygenesis, seeing

13
Language 28 (1952): 182-5. The question had already been broached by Colinet 1892 (see Schri-
jnen, KZ 38 (1905): 140) and Siebs 1904: 293-4.
14
Anatolian išḫ- never continues the prefix *es (see below). On the parallel change *sHK- > *sK-
(ὀλισθάνω – slide, Latin avis : Hitt. šuwaiš) see Peters Untersuchungen (1980): 71 fn. 34.
15
After Meillet 1929: 81; see M. Southern 1999: 134-8 and add R. Neuburger, IF 80 (1975) on
Latin abs-condere and ab-scondere.
16
Morpholog. Untersuch. IV (1881): 329f. fn. Compare Skr. ut-thita-, Greek ἐξ τοῦ > ἐκ τοῦ,
γέγραφ-σθαι > γέγραφθαι, *dek-s-to > Myc. de-ko-to etc.
17
Hitt. tiya- “step” beside ištanḫ- “try”, Ved. paśyati beside spáś- and tras beside stŕ̥bhis, Lat. tundo
beside studeo and the like.
92 George E. Dunkel

loss by degemination as the primary and accretion as a secondary source of s-mobile,

although s-addition is for Southern an analogic “sub-grammatical process” rather than
the trace of any specific morpheme.
At this point I would like to propose a specific new particle for the accretive aspect
of the synergism.
3. *es preceding roots. I suggest that both the augment *e- and the movable *s-
arose from an earlier particle of the form *es. Its original meaning was not preterital
but rather resultative, indicating that the verbal idea has been accomplished. Such a
meaning could, to be sure, easily be re-interpreted as preterital. However such a parti-
cle need not have been restricted to verbs.
After the present injunctive had been marked as “hic et nunc” with a postposed
near-deictic *-i, a corresponding resultative was created by prefixing *es-. Since its
meaning was neither “ibi et tunc” nor “olim” but rather something like “confectum”
or “completely, totally”, it could also be prefixed to aorist injunctives even without a
corresponding “hic et nunc”.
Present Aorist
Injunctive (timeless) *pe-i̯e-t *pē-s-t
hic et nunc *pe-i̯e-t-i
Accomplished *es pe-i̯e-t *es pē-s-t
After univerbation the boundary was shifted one phoneme left-ward on the model
of roots with non-movable *sK-;18 this engendered both the augment and the s-
allomorph simultaneously:
Metanalysis *e-spe-i̯e-t *e-spē-s-t
Redistribution *spe-i̯e-ti
Once the new allomorphs with initial *s- became prestigious, the alternation could
spread to non-verbal forms as well.
Unclear by this scenario remains why all roots do not show movable *(s)-: why are
there no *(s)dʰeh₁-, (s)bʰer-, (s)ĝneh₃-, (s)leik-, why *(s)pe- but no *(s)pre-? This problem
holds for any prefix theory which is semantically explicit; as long as the meaning is left
vague, one can simply assert that in roots with no movable *-s the combination did
not occur. Given the inescapable reality of prefixation at some stage, the assumption
of hyperforms after the metanalysis is a reasonable price to pay for this particle’s ex-
planatory advantages.
Excursus: Movable *s- among the other morpheme-classes. This topic has
never been examined systematically and would repay further study.

18
Cf. also e.g. *es sed- > *esed- > *e-sed-.
 The Indo-European Resultative Particle *es 93

a. Without being explicitly so called, movable *s- is often posited in suffixes in histori-
cal grammars of the IE dialects to explain otherwise irregular developments such as a
seemingly pre-resonantal loss of velars in Latin19 or aspiration of stops in Greek.20
Furthermore a variation between nominal derivatives in *-ko- and *-sko- is occasionally
encountered, as in Oscan toutiks versus Gothic þiudisks.21
The semantically empty presuffixal *-(s)- probably arose through metanalyses in de-
rivatives of s-stems (cf. Old Latin IOVXMENTA with iūgera) or of roots in final *-s (cf.
Latin tēla “cloth” < *teks-lā but tēlum “spear” < *tend-slo-).
b. While movable *s- has to my knowledge never been alleged among the nominal
endings, in the second singular perfect an allomorph *-sth₂e beside the usual *-th₂e has
repeatedly been proposed.22 This more likely originated by contamination with the
active ending *-s in the same person23 than (with Jasanoff) as a post-dental allomorph,
since no s-allomorphs exist for other dental-initial morphemes such as the endings *-ti
and *-to or the suffixes *-tó-, *-tér-, or *-tero-.
In Greek an intrusive -σ- occurs in various mediopassives, as in the plural endings
-µε(σ)θα and -σθε and before the ending in perfects like ἔγνωσµαι, πέφασµαι.
On *-(e)s following verb-endings see §8.
c. Among the particles movable *s- is seen in *(s)up “above/below”, -(s)ke “and” and
*su̯ō/-u̯e “like”. Historically, loss of etymological *s- is likely in *-ke from *sk-o “fol-
lows” and in comparative *-u̯e, whether due to allegro-treatment or to degemination
after final *-s. A morpheme-boundary shift in e.g. *ab-ske > abs-ke simultaneously
generated “adverbial *-s”.24 Meillet 1929: 81 saw metanalysed adverbial *-s initially in
*(s)up and *(s)h₂en- “without”.
d. In short: movable *s- in suffixes arose due to metanalytic gain and in particles due
to metanalytic loss and degemination, just as proposed above for roots. In IE and
Greek medio-passive endings, on the other hand, it originated by contamination with
other endings.
We now turn to avatars of resultative *es following roots.

19
In Latin the loss of velar before sR is followed by the īdem-lengthening: agmen–exāmen, sagmina–
sāmen, flagrāre–flamma < flāma, etc.
20
Cf. αἰχµή (Myc. a₃-ka-sa-ma) and τέχνη with *-smā and *-snā, as opposed to e.g. πυγµή and
ψίγνη, respectively.
21
While in verbal *-so- the s is not mobile.
22
C. Watkins 1962: 84-8; W. Cowgill, Evidence for Laryngeals (ed. W. Winter 1965): 172-3; G.
Schmid, Glotta 63 (1985): 90f.; J. Jasanoff FS Hoenigswald: 179, 1988a: 71; S. Insler, FS Strunk:
93-102.
23
In the mediopassive, Greek and Indic *-thēs < *-th₂e+es seems to combine the same two endings
in the reverse order; see however §8d.
24
See Anusantatyai (FS Johanna Narten), ed. E. Tichy, A. Hintze (2000): 17ff., also for earlier ideas
on adverbial *-s.
94 George E. Dunkel

4. Prehistory of the s-aorist. A. Meillet 1908 argued for a relatively late origin of
this category, as shown by its utterly anomalous (for athematics) lack of suffixal ablaut
and the atypical radical long-grade in the active.
Subsequent research has identified the ablaut-type as acrostatic and explained the
long-grade by analogic “Aufstufung” (K. Strunk, FT Berlin: 499). The suffix was traced
to a third singular preterital ending *-s reconstructed by C. Watkins on the basis of
Hittite,25 Tocharian, and Vedic26 material. He then described the subsequent metana-
lysis of the ending as a suffix and the construction of a paradigm on the basis of this
form (1962 passim), thus justifying the acrostasis.
Watkins derived the crucial third singular in *-s from a metanalysed root-extension,
comparing e.g. the Rgvedic hapax third singular aorist apās with the Hittite root pāš-.27
The root-extension itself arose through a metanalysis of the animate nominative end-
ing: *pre-s “asking (occurred)” (1962: 105f.). The full sequence of metanalyses was
thus, according to Watkins: nominative singular animate ending > root-extension >
third singular preterite ending > s-aorist suffix.
These ideas are not even mentioned by most of the recent and otherwise well-
informed introductions to IE linguistics,28 perhaps because the distribution of these
morphemes is as a whole neither overlapping nor complementary, as a structuralist
would expect. Watkins’ observations nonetheless remain a valid starting-point for a
“structural restatement” of the origins of these three anomalies in postradical *-s.
J. Jasanoff 1988a accepted the fundamental nature of the third singular in *-s but
reconstructed a paradigm with *-s- limited to the third person active indicative and the
entire subjunctive. By “a lexical peculiarity of the root *pre- alone” (1988a: 61), an
aorist with *ó/e ablaut and the *h₂-endings was suppleted by the imperfect of an athe-
matic s-desiderative with ē/e ablaut,29 “since verbs meaning ‘ask’ often have redundant
desiderative morphology” (ibid.). This approach’s exotic presuppositions (e.g. the abo-
ve-mentioned aorist), complex analogobatics, and its dependence on the single root
*pre- have encouraged the exploration of various alternatives.30
On Hittite -iš, an allomorph of -š, and its Phrygian equivalent -es, to which none of
the above scholars attach any importance, see §7e.
25
This is preferable to taking Hittite third singular preterital -š as analogically generalised from the
second person due to the merger (as -ta) between second singular *-th₂e and third singular *-t
(Kuryłowicz 1964: 157; on the final “prop-vowel” Melchert 1994: 175).
26
RV aprās “filled” (RV 10x, aprāt never), apās “drank” (only 5.29.8 beside apāt 4x), cf. third plural
dhās+ur beside dhāt.
27
Watkins 1962: 98-102, 1969: 53f. Hittite au(š)-, cited as a parallel (1969: 54), in fact shows pre-
cisely the reverse development, the ending -š here having secondarily become part of the root
(differently 1962: 78-9).
28
With the exception of Szemerényi 1996: 282 with fn. 26 (who rejects it).
29
Jasanoff’s separate theory of s-presents (1988b) must cope with the redefinition of Tocharian A
kñasäst as a preterite: Hackstein, FS Rix (1993): 151-8, Hardarsson, Wurzelaorist (1993): 76-9,
Schmidt-Winter, HS 105 (1992): 50-6, LIV2: 168-9 fnn. 3a-d.
30
See the honorand’s Selected Papers II: 570-2 as well as D. Ringe, MSS 51 (1990): 183-242 and D.
Adams 1994.
 The Indo-European Resultative Particle *es 95

I propose that the use of *-es and *-s as preterital endings and root-extensions and
of *-s as an aorist suffix arose through metanalyses of resultative *-es (with its aphaere-
tic post-vocalic allomorph *-s) after verbal roots. Through far simpler, this approach
explains much more material than any of the above.
5. Deictic order reversal. As a consequence of the IE word-structure formula
(P+) Word (+P), where Word can consist of either R(+S)+E or of simple P, certain
(but not all) particles could either precede or follow their heads. For this reason, vari-
ous particle-deictic combinations are found in both orders (cf. ∆ωρόθεος–Θεοδώρου,
Πατροκλεής–Κλεοπάτρα):
*e: Hitt. ki-nun – Lat. nun-c, Lat. ce-do – Gaul. du-ci
*i: Ved. -dr̥ś- – Gr. τοδ-, Lith. ĩ-pat – *pot-i
*u: Av. auu-at̰ – Ved. “adas” < *ad-au̯
*h₂u: Ved. u-tá – Goth. þa-u, Ved. u ca – co
*-ge: Gr. γάρ – Lith. ar̃gi, ar̃gu
The implication is that resultative *es, studied above preceding its head, could also
follow it.
6. The origin of athematic conjugation is generally seen in bare roots with ei-
ther e- or o-vocalism, the basis of the later “strong stems”. At first, unmarked forms
with the ending zero is thought to have indicated the functional zero-person. In the
imperative this is the second person,31 and such forms are in fact directly preserved:
Latin ī, es, Lith. eĩ-k, dúo-k.32 In the indicative, however, the zero-person is the third.
Such unmarked third-personal indicatives, originally indifferent to number and tense,
served as the basis for building paradigms (Watkins 1962: 90-6).
It was precisely such indifference that the addition of personal endings was in-
tended to counteract. At the start of EIE the two well-known ending-sets, active *-m -s
-t -ént and stative *-h₂e -th₂e -o -ér, were added to the *e- and *o-grade roots respectively.
The accented plural endings induced the radical zero-grade outside the singular.33
Much later came the primary markers, the differentiation into present/aorist, perfect
and mediopassive, and the augment.
7. *es following roots. Based on the preceding two sections I now suggest that *es
could be postposed to roots as well as preposed.
7a. In fact, postposition seems to have been the earlier practice, since it must have
taken place before any endings had been added on and a fortiori before quantitative

31
Watkins 1969: 51f., 119f., after Benveniste and Kuryłowicz.
32
On Lith. dúoki < *deh₃-Ø kid see §10a of Dunkel forthcoming.
33
On full-grade plural imperatives such as *h₁ei̯-te see MSS 46 (1985): 56 and §1b of Dunkel forth-
coming.
96 George E. Dunkel

ablaut, i.e. in pre-IE.34 At a stage without verbal or nominal endings it is a valid ques-
tion whether resultatives like *pe es “seeing accomplished” and *dʰoh₁ es “putting ac-
complished” can be characterised verbal or nominal at all. In the following I will as-
sume that both interpretations were possible and did in fact occur.
After univerbation and once quantitative ablaut came into effect, stress on one syl-
lable of these resultatives entailed a zero-grade in the other. The resulting forms were
then repeatedly metanalysed, whereby *-(e)s was taken either as a part of the root, or as
a suffix, or as an ending.
7b. Through morpheme-boundary loss *es sometimes became part of the root, i.e. a
root-extension. The resultative force was often lost due to semantic fading.35 Ac-
cented *-es extends zero-grade roots as in *u̯-es- “wear, have on” to *au̯- “wear
leather”,36 *h₂ĝ-es- (Lat. gero) to *h₂eĝ- “lead”,37 *ks-es- (ξέω “scratch”) to *kes- “arrange”
(FS Strunk p. 17), *zg-es- “quench” to *seg- “slow” (Lat. segnis),38 Gr. βδέω < *pzd-es-
to *pezd- (itself a lexicalised preverb-compound), and perhaps in Hitt. g(a)nešš- damešš-39
takešš- kallešš- as well.
Full-grade roots with the zero-grade extension *-s include *leu̯+s-, *h₂u̯eg+s-, and
*du̯ei̯+s- “hate”;40 S. Scarlata adds *dʰer+s- “dare” (cf. *dʰer- “hold firm”).
7c. The s-aorist stems arose from resultative syntagms in the same way: *pé es > *pé-
s-, with subsequent metanalysis of the ex-particle as a suffix plus a zero-ending of the
third singular as per Watkins. A similar metanalysis of words as stems is seen in pro-
nominal *toi̯-bʰis and *tās-ōm, with endings added to the fully characterised nominative
plurals, and in Luvian pluralising -nz-, based on the IE accusative plural.
7d. A similar (but earlier) metanalysis of resultative *es as a suffix, followed by the ad-
dition of inflectional endings, might be one source of the polygenetic nominal suffix
*-es-. The commonest subtype are the nomina rei actae such as *ĝenh₁es- “clan”, *nemes-
“handout”, *dees- “what is accepted”, *i̯eu̯ges- “what is yoked”, *h₃enes- “load”, *u̯ekes-
“word”, *menes- “mental agitation”, *h₃epes- “work”, *pees- “what is plucked”, etc. K.
Stüber has recently defined these as “resultative”.41 The double full-grade is of course
34
A stage of language without ablaut or the shibbolethic verbal or nominal endings is hard to
identify as IE at all, whence the term pre-IE. The addition of the characteristic endings was one
of the innovations which defines EIE.
35
See Scarlata, Wurzelkomposita im R̥g-Veda (1999): 527 fn. 729, last line.
36
Impossible for the LIV due to its self-imposed theoretical handcuffs (s.v. *h₂eu̯H- (p. 275)).
37
For inner-Latin proof of the reality of the initial laryngeal of *h₂eĝ- see 125 Jahre Indogermanistik in
Graz (ed. M. Ofitsch, Chr. Zinko 2000): 87-99.
38
*zges- is irrelevant for Siebs’ law because it does not contain the prefix, but rather the postposi-
tion.
39
On these see Jasanoff 1988b: 228ff. and Van den Hout, Ling. Happening … B. Schwarz (ed. Y.
Arbeitman 1988): 305ff. respectively.
40
Cf. *du̯ei̯- “fear” < *du̯o/i- “two” and Studies in Memory of Warren Cowgill (ed. C. Watkins 1987):
36.
41
K. Stüber, Die primären s-Stämme des Indogerman. (2002): 36, 224-31.
 The Indo-European Resultative Particle *es 97

no pre-IE archaism but rather the result of paradigmatic levelling, as generally agreed.
Formally, Root+*es was metanalysed as a locative with the ending zero. But the *es-
stems have other sources as well.
7e. Finally, postposed *-es functions synchronically as a personal ending of the zero-
person. Rather than, with Watkins, representing metanalysed root-extensions, these
arose parallel to them.
Hittite -iš is usually taken as an anaptyctic variant of the third singular -s seen in
dai-š pai-š nai-š au-š+ta etc. and Vedic aprā-s (Melchert 1984: 98, 108 and 1994 :174,
179f.).42 Now Phrygian -es (never written -is43 and hardly due to a Hittite substratum)
renders this improbable as it speaks for *-es. The regular development of final *-es in
Hittite is most likely -iš, cf. nepiš, Neša/Kaniš (Melchert 1994: 139); with *dʰoh₁ es > daiš
cf. *h₃eh₁-es-Ø > aiš “mouth”.
The third singular preterites in *-es of Hittite and Phrygian seem to preserve un-
changed the pre-IE full-grade root+*es. This of course cannot be; both Hittite daiš and
Phrygian edaes must be relatively recent re-creations of the old syntagm *dʰoh₁ es. This
implies that the free particle must have survived long enough to extend *dai (< *dʰoh₁-
e) and *eda respectively, cf. §8.
7f. In contrast to Watkins’ approach, considerable overlap can be observed between
different avatars of resultative *es.
One fifth of the LIV’s roots in movable *(s)- (29 of 152) form s-aorists, e.g.
*(s)tēg-s-, *(s)gēs-s-, *(s)lēh₂g-s-,*(s)nēubʰ-s-. The co-occurrence of pre- and postpositional
*es led to hypercharacterising contaminations such as:
*es pe- → *spe- *pé es > *pés-, analogic *pḗs-

*(s)pēs-
Nominal *-es-stems co-exist with other *es-avatars from the same root, cf. *u̯eĝʰ-es-,
*(s)teg-es- and *leu̯-es- with the s-aorists *u̯ēĝʰ-s-, *(s)tēg-s- and the extended root *leu̯-s-.
S. Scarlata compares, if only indirectly, the components of Hittite daiš, Phrygian
εδαες with those of the extended root *dʰeh₁-s- “god” (θεός, fēstus, fānum, Luvian tāna-
(Melchert, HS 110 (1997): 49) as reflecting a virtual *dʰeh₁ es (for imperatival θές <
*dʰh₁-es see §8e).
Such overlaps imply a single source for these anomalies.
42
The inherited distribution -iš after consonant, -š after vowel (dai-š “put”, da-š “took”, pai-š “ga-
ve” but akk-iš “died”, šakk-iš “knew”), was disturbed by analogies in both directions: ar-aš,
išpand-aš and ḫa-iš (to ḫa-mi “trust”). Watkins on the other hand related Hittite -iš to the perfect-
extension *-is(/er)- known from Latin (1969: 55-6, 155-7 after Meillet, Mélanges … Saussure
(1908): 99f.). On *-is- see further D. Adams 1994: 8-24 and S. Insler in FS Strunk (ajaniṣṭa < aja-
ni).
43
E.g. edaes/εδαες, eneparkes/ενεπαρκες, etoves, tiyes, εσταες, ποσεκανες; in Phrygian the postvocalic
allomorph -s has been replaced by -es.
98 George E. Dunkel

8. *es after complete verb-forms. Traces of resultative *es after verb-endings con-
tradict the chronology of §7a, but the assumption that free postpositive *es survived
into dialectal times is preferable to the usual welter of ad hoc explanations for the fol-
lowing material.
8a. In the third plural perfect Latin -ēr+unt and Hittite -er, -ir speak for *-ēr < *-er+s
while Vedic -ur and GAv. -ərəš continue Proto-Aryan *-r̥-š.44 “Resultative” here would
refer to the the state achieved by the subject, not (as in Wackernagel, Kleine Schriften:
1001) the object.
8b. The first plural endings45 *-me-s, *-me-n(e) and *-me-dʰh₂ seem to reflect differentia-
tions by various particles; in the second plural Vedic -tha+na stands beside Italo-Celtic
*-te+s (differently W. Cowgill, FT Regensburg: 59f.).
8c. An insular Celtic main-clause enclitic *(e)s in “Wackernagel position” has been
proposed for nearly a century by Celtologists who despair of explaining the difference
between absolute and conjunct inflections as reflecting the primary – secondary oppo-
sition.46 The present study provides comparative underpinnings for this hypothesis;
the tenacity of postpositive *es is a very conservative feature of these dialects.
8d. That the Graeco-Indic second singular mediopassive *-th₂e+es contains resultative
*-es (against fn. 23) also begins to seem more plausible in the present context; the con-
traction antedated the pre-Anatolian (Middle IE) laryngeal coloring of *e.
8e. S. Scarlata suggests that the enigmatic Greek second singular aorist imperatives δός
θές ἕς might continue *dh₃-és (etc.), replacing the earlier *déh₃-Ø (Lith. dúo-k: §6) just as
*h₁i-dʰí replaced *h₁ei̯-Ø.47
8f. The Old Persian third singular imperfects and thematic optatives in -š might belong
here as well.48
8g. The Oscan secondary third plural in -ns can however be traced back to *-nt just as
well as to *-nt+s, as shown by the neuter of the present active participle with -ns < *-
nt-Ø.
9. For IE *ĝʰ(þ)i̯és “yesterday” three etymological approaches have heretofore
been proposed.
9a. Common to most is K. Brugmann’s utterly unwarranted assumption that the initial
*ĝʰ- is near-deictic in function, for which he cited Indic ahám mahyám and Latin mihi

44
J. Jasanoff in Sound Law and Analogy (FS R. Beekes, ed. A. Lubotsky 1997): 120 and 1988a: 71;
skeptical Kümmel, Perfekt im Indoiran. (2000): 44.
45
Inclusive *me is opposed to exclusive *u̯e in the dual; more elsewhere.
46
Most recently by W. Cowgill, FT Regensburg: 40ff., FT Berlin: 109ff.
47
Differently K. Strunk, FS A. Dihle (1993): 486-72.
48
With Watkins 1962: 92-3; cf. Brandenstein-Mayrhofer, Handbuch des Altpers. (1964): 79, Kury-
łowicz 1964: 157.
 The Indo-European Resultative Particle *es 99

and hic (1904: 70-1). But even aside from the irregular aspirate of ahám, deictic use of
emphatic *ĝʰo/i is an innovation of Latin-Faliscan only, not a proto-linguistic feature,
especially since Melchert’s proof that it is not reflected by Hittite ka-, Luvian za-
(1994: 251f.). Nor is a zero-grade of this particle otherwise known.
A suffix-boundary is usually assumed before the *-i̯-, with what follows taken as ei-
ther adjectival *-i̯o-49 or as contrastive/intensive *-i̯es-.50
J. Schindler derives IE thorn from tautosyllabic *KT- as well as *TK-clusters, al-
lowing a pre-form*ĝʰ-di̯es (Die Sprache 23 (1977): 34). But the attractive comparison
with Vedic sadyás, adyá and Latin hodie, Faliscan foied nonetheless still presupposes, un-
acceptably, that the element *ĝʰ- was deictic.
J. Puhvel’s derivation of the thorn-cluster by metathesis from *dʰĝʰ- reveals for him
a contrast-formation to the preform behind Gothic dags and pre-Indic *(d)ahar-.51
However this is vitiated by Tocharian A ksär “morning”, assuming this reflects
*ĝʰes-tro- (cf. A tkaṃ with Greek χθών).
9b. I suggest that the original form for “yesterday” was simply *ĝʰes, cf. Lat. her-ī. This
could be extended by adverbial endings such as *-ter (cf. *-tr-o- in Lat. hes-ter-nus, Goth.
gis-tra-, Toch. A. ksär) and *-dʰi. Final accent or allegro-usage reduced *ĝʰez-dʰí to *ĝʰz-
dʰí52 (extended by adverbial -ς (cf. αὖθις) and then -δον in Greek χθιζόν).53 Persevera-
tion (homoioteleuton) from resultative syntagms added final *-es still in IE times:
*doh₃ es ĝʰz-dʰí > *doh₃ es ĝʰz-dʰi̯+és.
Cf. the transfer of a pseudo-augment to “yesterday” in Greek:
ἔλιπον χθές > ἔλιπον ἐ+χθές.54
Alternatively, *ĝʰzdʰés (of which Greek χθές and Vedic hyás are varying simplifica-
tions) may have arisen by contamination of *ĝʰzdʰí by *ĝʰés.
Perhaps even the simple *ĝʰes arose from *ĝʰ “behind” by the same perseveration.
Cf. in the reverse direction (i.e. from particle to verb) voluntative *h₁ei̯-oh₁ for *h₁ei̯-o-mi
after *eĝH-oh₁ and imperative *h₁i-dʰi for *h₁ei̯-Ø after *i-dʰi “here”.55
10. To summarise: The assumption of an IE particle *es with resultative value
makes possible a simpler interpretation of the three verbal *s-morphemes connected
by C. Watkins: root-extensions (*u̯-es-, *h₂ĝ-es-; *leu̯-s-, *dʰeh₁-s-), s-aorist stems
49
E.g. Brugmann 1904: 72; Specht, KZ 68 (1944): 201-4.
50
Specht ibid.; Puhvel, FS Hoenigswald: 317 = Epilecta: 45.
51
Op. cit. in the previous fn. (despite Schindler’s “unanalyzable”, Die Sprache 23 (1977): 32) after
M. Durante, Ricerche linguistiche 1 (1950): 248.
52
Here*-z- would block thorn – which is unneeded once *-dʰi is recognised.
53
Homeric χθιζά might be Arcadian to judge by Hesychian θύρδα.
54
As if he played it yesterday were to engender he played it yesterdayed. The free resultative postposition
would have occurred in the allomorph *-s.
55
More on perseveration in the forthcoming Indogermanische Syntax – Fragen und Perspektiven, ed. H.
Hettrich; for the delocatival imperative in *-dʰi see §11d of Dunkel forthcoming.
100 George E. Dunkel

(*(s)pē-s- *u̯ēĝʰ-s-), and third singular preterites (Hitt. daiš, Phryg. edaes); compare fur-
ther the nominal *-es-stems from many of the same roots (*leu̯-es-, *u̯ēĝʰ-es-). It also
accounts for both the augment and for movable *s- and offers a unified resolution of
various dialectal verb-forms in final *-(e)s. Finally, repeated perseverations allow deriv-
ing the IE word for “yesterday” thornlessly from *ĝʰ “behind”.
This is all possible due to the IE word-structure principles, which allowed this par-
ticle to be pre- as well as postposed and thus to be metanalysed in various ways. Al-
though resultative *es may seem extremely protean, there are clear limits to its applica-
bility, i.e. I see no connection between it and the following s-endings: animate nomina-
tive singular, animate nominative plural, genitive singular, or the second singular active
of the *m-conjugation.

Repeated references:
Adams 1994: “The Tocharian Class III Preterite”, in In Honorem Holger Pedersen, ed. B. Nielsen and
J. Rasmussen 1994, 1-28
Brugmann 1904: Die Demonstrativpronomina der idg. Sprachen, in Abhandl. der philol.-histor. Klasse der kö-
nigl. sächs. Gesellschaft der Wissenschaften XXII.vi
Dunkel forthcoming: “The deictic origin of the Greek ka-aorist and ka-perfect”, in Indo-European
Word Formation – Inventory and Analysis, ed. B. Olsen
FS Hoenigswald: Festschrift for Henry Hoenigswald, ed. G. Cardona and N. Zide, 1987
FS Strunk Verba et structurae, ed. H. Hettrich et all., 1995
FT Berlin: Grammatische Kategorien, ed. B. Schlerath 1985
FT Regensburg: Flexion und Wortbildung, ed. H. Rix 1975
Hirt 1927: Indogermanische Grammatik I
Jasanoff 1988a: “The sigmatic aorist in Tocharian and Indo-European”, in TIES 2, 62-76
Jasanoff 1988b: “PIE *ĝnē- ‘recognise, know’”, in Die Laryngaltheorie, ed. A. Bammesberger, 227-39
Kurylowicz 1964: Inflectional Categories of Indo-European
LIV Lexikon der idg. Verben, ed. H. Rix, 2. Auflage 2001
Mayrhofer 1986: Segmentale Phonologie des Indogerman. (= Indogerman. Grammatik, ed. J. Kurylowicz,
Band I/2)
Meillet 1929: “Sur latin sub, super”, BSL 30, 80-1
Melchert 1994: Anatolian Historical Phonology
Siebs 1904: “Anlautstudien”, KZ 37, 277-324
Southern 1999: Sub-grammatical Survival (JIES Monograph 24)
Strunk 1994: “Der Ursprung des verbalen Augmentes”, in Bopp-Symposium, ed. R. Sternemann, 270-
84
Szemerenyi 1996 Introduction to Indo-European Linguistics, 5th edition
Watkins 1962: IE Origins of the Celtic Verb
Watkins 1969: Geschichte der indogerman. Verbalflexion (= Indogerman. Grammatik, ed. J. Kurylowicz,
III/1)